Sliding-Window Peak Demand Control for EV Depot Charging
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current methods for managing peak power demand at power infrastructure sites, such as EV depots, are inadequate due to uncertainty and scalability issues, often resulting in penalties or inefficient resource allocation, as they rely on historical estimates and fixed peak power settings that do not account for future uncertainties.
Innovation Solution
A method using a sliding time window to determine optimal peak power demand by combining past, current, and forecasted values, allowing for real-time adjustments based on historical data, operating configurations, and probability distributions, to set an accurate peak power demand for future billing periods.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of operation
If fixed peak power demand settings based on historical estimates are used, then device complexity is reduced and ease of operation is improved, but measurement precision of future peak power demand deteriorates and reliability of power management worsens
Solution Approach 1:
The system transitions from static fixed peak power settings to dynamic adjustment mechanisms. The operator device continuously receives real-time power consumption data, updates predictions using machine learning models, and adjusts peak power demand settings dynamically to adapt to changing conditions while maintaining operational simplicity.
Solution Approach 2:
The system implements closed-loop feedback where actual power consumption data is continuously monitored, compared against predictions, and used to refine future peak power demand estimates. This feedback mechanism improves prediction accuracy over time without increasing operational complexity for the user.
2Device complexity
If historical estimates and fixed settings are used for peak power demand, then device complexity is reduced, but adaptability to future uncertainties and reliability of power management deteriorate
Solution Approach 1:
The system performs self-adjustment through automated machine learning models that continuously learn from historical and real-time data. The operator device automatically updates predictions and adjusts peak power demand settings without requiring manual intervention or complex configuration, maintaining simplicity while improving adaptability.
Solution Approach 2:
The system proactively predicts future peak power demand before it occurs using trained machine learning models. By preparing predictions in advance based on historical patterns and real-time trends, the system adapts to future uncertainties before they manifest, avoiding the need for complex reactive adjustments.
3Productivity
If maximum power equipment capacity is used early as possible, then productivity of charging operations is improved, but loss of energy increases due to unnecessary peak power consumption
Solution Approach 1:
Instead of always using maximum equipment capacity, the system applies partial action by adjusting peak power demand to match actual needs. The machine learning models predict optimal charging schedules that utilize equipment capacity efficiently without consistently operating at maximum levels, reducing energy waste while maintaining adequate productivity.
4Device complexity
If conventional estimation methods are used for peak power demand, then device complexity is reduced, but manufacturing precision of power management decisions and reliability deteriorate
Solution Approach 1:
The system replaces manual estimation methods with automated machine learning-based prediction systems. The operator device automatically collects data, trains models, and generates predictions without requiring manual analysis or complex configuration procedures, improving decision precision while keeping the user interface simple.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
Operators typically utilize techno-economic means to set peak power demand in power infrastructure sites, such as power depots and microgrids. However, conventional means tend to either produce sub-optimal values of peak power demand or be too computationally expensive to be performed in a real-time or scalable manner. Accordingly, disclosed embodiments utilize a sliding time window to continuously or periodically determine peak power demand in past, current, and future portions of a current time period. These embodiments are able to determine an optimal peak power demand for the current time period, while remaining computationally feasible for real-time performance and being scalable with the complexity of optimization.